Unmanned aerial vehicle agricultural condition monitoring aerial survey method based on space sampling optimization

Abstract

The invention belongs to the technical field of agricultural condition monitoring, and particularly relates to an unmanned aerial vehicle agricultural condition monitoring aerial survey method based on space sampling optimization, which comprises the following steps: constructing a digital farmland base model, and dividing minimum agricultural units with independent agricultural attributes; the number of sampling points of each unit is calculated based on a sampling model, Monte Carlo random sampling is executed, abnormal points are eliminated in combination with a space coordinate verification mechanism, and dynamic sampling planning is achieved; generating an intelligent route by adopting a traveling salesman problem optimization algorithm; the aerial survey task is automatically controlled by calculating the flight height; and dynamically adjusting the sampling density according to the crop growth heterogeneity coefficient to obtain the agricultural condition information of the minimum agricultural unit, and integrally generating a plot-level monitoring report. According to the unmanned aerial vehicle agricultural condition monitoring aerial survey method based on space sampling optimization, the efficiency and precision of farmland monitoring can be remarkably improved through the collaborative design of the multi-target ant colony algorithm optimization air route and the dynamic stratified sampling mechanism, and accurate decision support is provided for farmland management.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial survey and agricultural information technology, and relates to an unmanned aerial survey method for agricultural condition monitoring based on spatial sampling optimization, and specifically to an unmanned aerial vehicle agricultural condition monitoring method based on spatial sampling optimization and intelligent path planning, which is suitable for farmland growth assessment, pest and disease monitoring and precision agricultural management. Background Art

[0002] With the rapid development of drone technology, its application in the field of agricultural monitoring has become increasingly widespread, bringing significant changes to modern agriculture. Due to its efficient, accurate and real-time monitoring capabilities, drones have become an indispensable and important tool in modern agricultural production, and are widely used in tasks such as farmland pest and disease monitoring, crop growth monitoring, and soil moisture monitoring. Drones are equipped with a variety of remote sensing sensors, such as high-definition digital cameras, multispectral sensors, thermal imaging sensors, etc., to obtain multi-dimensional and high-precision farmland information, including plant number, vegetation coverage, leaf area index, biomass, plant height, nitrogen nutrition status, water status, etc. These parameters together characterize the growth of crops and are ultimately related to yield.

[0003] However, most of the existing drone agricultural data collection and monitoring technologies rely on experimental methods in the field of scientific research, and their maturity and applicability are not yet sufficient to meet the needs of large-scale commercial applications. Although the existing drone agricultural monitoring technology can plan routes based on the set flight altitude, track overlap rate and flight speed, this method still has many defects in actual farmland management.

[0004] First: Existing technology requires drones to cover the entire farmland area. During the flight, the drones plan the aerial survey route based on factors such as the survey area, recognition accuracy, resolution of the airborne imaging equipment, flight altitude, and route overlap rate, and obtain large-scale impact data according to the set route spacing (D) and exposure spacing (B). After these image data are transmitted back to the ground station, complex operations such as image stitching, three-dimensional point cloud computing, and orthophoto generation are required to obtain orthophotos of the monitored plots. This process places extremely high demands on data transmission, computer storage space, and computing performance, resulting in poor timeliness of agricultural monitoring and high technical barriers, making it difficult to adapt to the agricultural monitoring needs of actual agricultural production.

[0005] Second: When higher monitoring accuracy is required, the flight altitude needs to be adapted to factors such as the focal length of the airborne imaging equipment, the size of the image sensor, and the sensor resolution (pixel scale). However, the route planning of the existing technology is mainly aimed at the complete area of ​​​​the farmland. In order to ensure the success rate of image stitching, the heading overlap rate is generally required to be ≥65%, and the lateral overlap rate is generally required to be ≥35%. Under the premise of a lower flight altitude, the area of ​​​​the sensor monitoring area is greatly reduced, which will greatly increase the flight route and image acquisition density, resulting in excessive flight time and low flight efficiency. Therefore, there is a prominent contradiction between monitoring accuracy and monitoring efficiency in the existing methods. The growth of crops is often a dynamically changing process, and obtaining data by flying in a way that fully covers the entire farmland area not only wastes a lot of flight time, but also increases the complexity and time cost of later data processing. The route planning method of the existing technology is obviously unable to meet actual needs.

[0006] Third: There is a lack of monitoring methods for spatial management units with independent agronomic management characteristics. For specific minimum agricultural units, the determination of sampling points is often done manually. Whether the number of sampling points can reflect the actual spatial heterogeneity, whether there is redundancy in the sampling points, and the lack of efficient flight route planning and control methods for the sampling point set in a single operation limit the feasibility of high-efficiency drone monitoring of each minimum agricultural unit in the field based on statistical concepts.

[0007] For field production, especially in farmland management decisions, the overall growth information of farmland does not require too high spatial resolution, but more attention is paid to the accuracy and timeliness of monitoring. The existing agricultural data collection methods based on drones have a contradiction between monitoring accuracy and monitoring efficiency, and cannot simultaneously meet the requirements of high accuracy and high timeliness of agricultural data collection. Therefore, how to improve the accuracy and timeliness of agricultural data collection and monitoring in agricultural production is a technical problem that technicians in this field need to solve urgently. Summary of the invention

[0008] In view of the above technical problems, the present invention provides a method for agricultural condition monitoring and aerial surveying by unmanned aerial vehicles (UAVs) based on spatial sampling optimization, which achieves the technical effect of reducing the number of image returns and computer processing pressure, and improving the accuracy and timeliness of agricultural condition data collection.

[0009] In view of this, the present invention provides a method for agricultural condition monitoring and aerial surveying by unmanned aerial vehicle based on spatial sampling optimization, comprising the following steps:

[0010] A method for agricultural condition monitoring and aerial survey using a drone based on spatial sampling optimization includes the following steps:

[0011] S1: Pre-build the GIS model of the monitoring plots, and draw the plot boundaries and the topological structure of the minimum agronomic unit based on the geographic coordinate information;

[0012] S2: Establish a sampling model based on the agricultural monitoring accuracy index, and calculate the optimal number of sampling points for each minimum agronomic unit based on the minimum agronomic unit area parameter and crop growth heterogeneity index;

[0013] S3: Execute the Monte Carlo random sampling algorithm to generate a set of monitoring sampling points that meet the preset number and conform to the spatial distribution constraints, and exclude abnormal sampling points that exceed the unit boundary or have overlapping areas;

[0014] S4: Apply the traveling salesman problem optimization algorithm to generate global route planning, with the minimum number of flight turns, the shortest total flight distance and the consistent image acquisition azimuth as multi-objective optimization constraints, and intelligently generate drone routes;

[0015] S5: Implement automated control of aerial survey tasks, collect fixed-point images at preset altitudes based on RTK positioning technology, and build a metadata indexing system with temporal and spatial correlation;

[0016] S6: Deploy a deep learning image analysis model, analyze crop parameters through feature extraction technology, use sampling statistics to obtain the smallest agronomic unit information, and integrate to generate plot-level monitoring reports.

[0017] Furthermore, in step S1, a digital farmland base model with centimeter-level accuracy is constructed through spatial registration of satellite remote sensing images and drone aerial survey orthophotos.

[0018] Furthermore, the minimum agronomic unit is a spatial management unit with independent agronomic management characteristics.

[0019] Furthermore, the division method of the minimum agronomic unit includes at least one of an irrigation system topology structure and an agricultural machinery operation partition.

[0020] Furthermore, in step S2, the sampling model of the sampling point satisfies ,in n is the number of sampling points, Z is the confidence coefficient, p is the estimated parameter variance, e is the allowable error range.

[0021] Furthermore, a spatial coordinate verification mechanism is established. In step S3, Monte Carlo random sampling is performed and combined with the spatial coordinate verification mechanism to eliminate abnormal points.

[0022] Furthermore, in step S4, an improved ant colony optimization algorithm is used, the pheromone volatility coefficient ρ=0.1, the inspiration factor α=1, the expectation factor β=5 are set, a multi-objective optimization function is constructed, and the multi-objective traveling salesman problem is solved, wherein the multi-objective optimization function is:

[0023]

[0024] is the weight coefficient, is the node spacing, is the number of turns, To preset the optimal collection azimuth, keep it parallel to the sowing row azimuth.

[0025] Furthermore, in step S4, the improved ant colony algorithm introduces a quantized pheromone attenuation mechanism. When the path pheromone concentration exceeds a threshold When Dynamic reset, λ is the decay rate, and t is the number of iterations.

[0026] Furthermore, in step S5, the method for determining the preset flight altitude includes:

[0027] Based on the spatial resolution requirements of the target agricultural parameters, the formula Calculate the flight altitude, where:

[0028] H is the flight altitude of the drone, in meters;

[0029] GSD is the ground sampling distance, in centimeters / pixel, which is set according to the monitoring accuracy requirements. GSD ≤2D, D is the minimum recognition size of the target crop feature;

[0030] f is the focal length of the camera, in millimeters;

[0031] p x is the camera pixel size in microns;

[0032] The altitude control error does not exceed ±0.5%H.

[0033] Furthermore, the principle of the sampling statistical method in step S6 is:

[0034] Spatial stratified sampling is performed for each minimum agronomic unit, and the heterogeneity coefficient of crop growth within the unit is used to determine the spatial distribution of the crop growth. d = s / m , s is the standard deviation of historical data, m As the mean, the sampling density is adjusted dynamically:

[0035] when d ≥0.3, high-density sampling mode is adopted, and the number of sampling points n ≥5;

[0036] When 0.1≤ d<0.3, medium density sampling mode is adopted, 3≤ n <5;

[0037] when d When <0.1, a low-density sampling mode is used. n =1;

[0038] The sampling process verifies the representativeness of the sample through spatial autocorrelation analysis to ensure the Moran index I ∈[-0.2,0.2].

[0039] Compared with the prior art, the drone agricultural monitoring aerial survey method based on spatial sampling optimization described in the present invention has the following advantages:

[0040] (1) The drone agricultural monitoring and aerial survey method based on spatial sampling optimization described in this application divides the farmland into multiple minimum agronomic units by using a sampling strategy, sets sampling points in each operation unit according to needs, and plans the flight route based on these sampling points, thereby reducing redundant data collection and significantly improving flight efficiency. As a result, the drone can fly along the optimal path, reduce unnecessary flights and sampling, optimize the data collection process, and ensure that accurate and representative farmland data can be obtained while improving monitoring efficiency.

[0041] (2) The UAV agricultural monitoring aerial survey method based on spatial sampling optimization described in this application solves the contradiction between improving the UAV monitoring accuracy and improving the monitoring efficiency by optimizing the route using a multi-objective ant colony algorithm and co-designing the dynamic stratified sampling mechanism. It is suitable for rapid and precise agricultural monitoring of field crop scenes. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a schematic diagram of the smallest agronomic unit and the UAV route planning in the embodiment of the present invention;

[0043] The symbols in the figure are:

[0044] Farmland area 1, shelterbelt 2, road 3, minimum agronomic unit 4, sampling point 5, flight route 6. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.

[0046] In the description of the present application, it should be noted that the terms used herein are only for describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present application. For ease of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship. The technology, methods and equipment known to ordinary technicians in the relevant field may not be discussed in detail, but in appropriate cases, the technology, methods and equipment should be regarded as part of the authorization specification. In all examples shown and discussed here, any specific value should be interpreted as merely exemplary, rather than as a limitation. Therefore, other examples of the exemplary embodiments may have different values. It should be noted that similar numbers and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.

[0047] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0048] It should be noted that, in the description of the present application, the orientation or positional relationship indicated by terms such as "front, back, up, down, left, right", "lateral, vertical, perpendicular, horizontal" and "top, bottom" are usually based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description. Unless otherwise stated, these orientation words do not indicate or imply that the device or element referred to must have a specific orientation or be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the scope of protection of the present application; the orientation words "inside and outside" refer to the inside and outside relative to the contour of each component itself.

[0049] It should be noted that, in the present application, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises one..." does not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be noted that the scope of the method and device in the embodiment of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0050] The embodiments of the present application are described above in conjunction with the accompanying drawings. In the absence of conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

[0051] Existing drone agricultural monitoring technology usually needs to cover the entire farmland area, obtain a large amount of data through frame-by-frame images, and perform complex stitching and three-dimensional point cloud construction to generate orthophotos. This process requires extremely high computer storage and processing performance, and the stitching and data processing time is long, resulting in poor monitoring timeliness. In addition, the route planning of existing technologies is usually based on the complete area of ​​​​the farmland, the flight time is too long, the efficiency is low, and it is difficult to meet the needs of actual farmland management. Especially in field production, monitoring does not require too high spatial resolution, but pays more attention to accuracy and timeliness. According to the heterogeneity of field crop growth, a statistical sampling model is established, which can achieve the mean result of the monitoring area based on limited sampling point monitoring.

[0052] The present application discloses a method for drone agricultural monitoring based on spatial sampling optimization, comprising:

[0053] S1: Pre-build the GIS model of the monitoring plot, draw the plot boundary and the topological structure of the minimum agronomic unit based on the geographic coordinate information, and establish the spatial planning benchmark;

[0054] S2: Establish a sampling model based on the agricultural monitoring accuracy index, and calculate the optimal number of sampling points for each minimum agronomic unit based on the minimum agronomic unit area parameter and crop growth heterogeneity index;

[0055] S3: Execute the Monte Carlo random sampling algorithm to generate a set of monitoring sampling points that meet the preset number and conform to the spatial distribution constraints, and establish a spatial coordinate verification mechanism to exclude abnormal sampling points that exceed the unit boundary or have overlapping areas;

[0056] S4: Apply the Traveling Salesman Problem (TSP) optimization algorithm to generate global route planning, with the minimum number of flight turns, the shortest total flight distance, and the consistent image acquisition azimuth as multi-objective optimization constraints;

[0057] S5: Implement automated control of aerial survey tasks, collect fixed-point images at preset altitudes based on RTK positioning technology, and build a metadata indexing system with temporal and spatial correlation;

[0058] S6: Deploy a deep learning image analysis model, analyze crop parameters through feature extraction technology, use sampling statistics to obtain the smallest agronomic unit information, and integrate to generate plot-level monitoring reports.

[0059] In farmland management, it is not necessary to conduct comprehensive coverage sampling of the entire farmland. Traditional methods often lead to low monitoring efficiency and long processing time due to the huge amount of data. The drone agricultural monitoring aerial survey method based on spatial sampling optimization disclosed in this application includes the synergy of steps such as geospatial modeling, dynamic sampling planning, intelligent route generation, precise altitude control, and adaptive sampling execution. The farmland area 1 boundary and the minimum agronomic unit 4 topology are accurately constructed through the geographic information system model to provide a benchmark for subsequent spatial planning. According to the accuracy requirements of agricultural monitoring, combined with the area of ​​the minimum agronomic unit and the heterogeneity of crop growth, the optimal number of sampling points 5 for each minimum agronomic unit is scientifically calculated to ensure that the sampling is both representative and not excessive. Then, the Monte Carlo random sampling algorithm is used to intelligently generate a set of sampling points that meet the spatial distribution constraints, and the spatial coordinates are used. The verification mechanism removes abnormal points to ensure the accuracy of sampling. Then, the traveling salesman problem optimization algorithm is used to generate an efficient global flight route 6 plan with the goal of reducing the number of flight turns, shortening the total flight distance, and maintaining the consistency of the image acquisition azimuth. During the aerial survey of the planned flight route 6, the UAV uses RTK positioning technology to achieve precise fixed-point image acquisition, and establishes a time-space associated metadata index system to facilitate subsequent data processing. Finally, through the deep learning image analysis model, the collected image data is feature extracted and agricultural parameters are analyzed, and the sampling statistical method is used to obtain the agricultural information of the smallest agronomic unit, and finally integrated to generate a plot-level monitoring report to achieve accurate monitoring of farmland conditions.

[0060] The drone agricultural condition monitoring and aerial survey method based on spatial sampling optimization described in the present application reduces the monitored object to the smallest agronomic unit, samples based on the smallest agronomic unit and performs optimized route planning. The optimized route planning can accurately control the selection of sampling points, ensure that representative data of each smallest agronomic unit is collected, avoid the generation of too much useless data, thereby reducing the pressure of data feedback and subsequent processing, significantly improving flight efficiency and the timeliness of data collection, greatly improving the efficiency and accuracy of farmland monitoring, providing accurate decision-making support for farmland management, greatly improving the efficiency of agricultural condition monitoring, and meeting the needs of modern agriculture for fast and accurate information.

[0061] As a preferred example of the present application, in step S1, when drawing the boundary of the farmland area, the satellite remote sensing image and the drone aerial survey orthophoto image are spatially registered, and the polygonal boundary is drawn by manual hand-drawing or geographic information system software to construct a digital farmland base model with centimeter-level accuracy. In the example of the present application, the farmland area 1 generally refers to a whole piece of farmland, which is generally surrounded by a protective forest 2 or a road 3 as an isolation. Because a piece of farmland is large in area, agronomic measures such as irrigation, fertilization, and spraying cannot be carried out at the same time, thus generating the minimum agronomic unit 4 in the farmland. When determining the minimum agronomic unit 4, it is first necessary to draw the border of the farmland area 1. You can use a satellite map with geographic coordinates, or an orthophoto map with geographic coordinate information synthesized by a drone. When drawing the farmland boundary, you can use hand-drawing on the remote control, or you can use geographic information system (GIS) software, such as ArcGIS, to accurately draw the polygonal boundary using shp files. This application uses satellite maps or orthophotos to draw farmland boundaries, which not only improves the accuracy of drawing, but also can quickly obtain geographic coordinate information, thereby providing a reliable basis for subsequent flight planning and data collection. With the support of professional software such as ArcGIS, the boundaries of the smallest agronomic units can be accurately drawn and numbered, making the management of the entire farmland area more scientific and efficient; at the same time, the flexibility of hand-drawing or using GIS tools also greatly improves the convenience and adaptability of operation, so that farmland management in different areas can be flexibly adjusted according to actual conditions.

[0062] As a preferred example of the present application, the minimum agronomic unit is a spatial management unit with independent agronomic management characteristics, such as an area where irrigation, fertilization, spraying and other operations can be performed independently. By dividing the farmland into multiple such minimum agronomic units, each operation unit can be independently managed and monitored in a refined manner. For example, crops in different operation units may have different growth states, so targeted agronomic measures can be taken in each operation unit, such as adjusting the amount of irrigation or fertilization according to the needs of the crops. This division method not only takes into account the actual use of the farmland, but also provides greater flexibility in farmland management.

[0063] By dividing the farmland into the smallest agronomic units, this application enables each operation unit to independently carry out agronomic measures, making farmland management more refined and scientific. Each operation unit can independently collect data and conduct agricultural condition monitoring, avoiding the data redundancy caused by large-scale monitoring of the entire farmland area in traditional methods. This not only ensures the accuracy and timeliness of the monitoring data, but also reduces unnecessary data processing work and improves the speed and accuracy of data analysis.

[0064] As a specific example of the present application, the monitored plots are divided into a set of minimum agronomic units with independent agronomic attributes according to the topological structure of the irrigation system and the agricultural machinery operation partitions to form a GIS topological structure. In the example of the present application, the minimum agronomic unit can be divided according to the energy efficiency of the agronomic operation. For example, a plurality of wells are set in the area where the farmland is located, and the minimum agronomic unit is defined according to the maximum operating area of ​​each well working at the same time, or the minimum agronomic unit is defined according to the operating capacity of the fertilizing machinery at each preset time (for example, 2h, 4h, 8h, 16h, 24h, etc.). By dividing the farmland into irrigation, fertilization, spraying or mechanical operation partitions, each partition can be independently operated and managed according to the different needs of crops, which greatly improves the resource utilization efficiency and the pertinence of the operation. For example, in the irrigation partition, the drone can monitor the soil moisture according to the actual water demand of each area, and in the fertilization management partition, the amount of fertilizer and the method of fertilization can be adjusted according to the nutrient status of the soil. This fine division of operation units not only improves the efficiency of farmland management, but also reduces the waste of agronomic measures. After the drone conducts precise monitoring based on these partitions, it can provide more accurate agricultural data, helping farmland managers to understand the crop growth status of each operation unit in real time and adjust agronomic measures in a timely manner. In the example of this application, the division of the minimum agronomic unit under different agronomic measures can be the same or different. There are differences in the province and city where the farmland is located and the type of farmland crops. The division method of the minimum agronomic unit is different, but it is generally based on the convenience of farmland management decisions. A piece of farmland is divided into many small plots of similar size and different shapes. This is called the minimum agronomic unit.

[0065] Through the above-mentioned precise zoning method, the present application allows drones to conduct targeted monitoring in each operation unit, collect agricultural data in each zone, ensure that agronomic measures in each area are reasonably implemented and receive timely feedback, and help farmland managers understand the crop growth status of each operation unit in real time and adjust agronomic measures in a timely manner. This not only improves the accuracy and timeliness of farmland monitoring, but also ensures the intelligence and efficiency of farmland management.

[0066] As a preferred example of the present application, in step S2, the number of sampling points in the minimum agronomic unit is dynamically determined according to the area of ​​the minimum agronomic unit and the uniformity of crop growth. In the example of the present application, after the farmland area 1 and the border of the minimum agronomic unit 4 are determined, the divided minimum agronomic unit 4 is numbered, and a unique number is assigned to each minimum agronomic unit according to the position of the minimum agronomic unit 4, as well as the area of ​​the minimum agronomic unit and the uniformity of crop growth, and the number of sampling points to be collected in each unit is dynamically determined. In this process, the selection of sampling points follows the principle of uniform distribution, and the sampling points can be randomly generated. The sampling point data that is not close to the border of the minimum agronomic unit is selected, and the specific geographical coordinates of each sampling point are determined by precise calculation to ensure that the sampling points can fully and evenly reflect the agricultural conditions in the operation unit.

[0067] This application avoids the monitoring blind spots that may be caused by traditional fixed sampling points and improves the utilization efficiency of monitoring resources by dynamically determining the number of sampling points within the minimum agronomic unit. The uniform distribution of sampling points ensures that the monitoring data can fully reflect the crop growth conditions within the operation unit, and improves the accuracy and reliability of the monitoring results. At the same time, the monitoring route planning strategy of the present invention does not increase the amount of monitoring data geometrically due to the improvement of monitoring accuracy. The number of sampling points is only determined by the spatial heterogeneity of the field. It can fundamentally solve the contradiction that the accuracy and efficiency of drone monitoring cannot be improved synergistically, give full play to the inherent advantages of drones with high spatiotemporal resolution, and accelerate the application of drone monitoring.

[0068] As a preferred example of this application, in step S2, the sampling model of the sampling point satisfies ,in n is the number of sampling points, Z is the confidence coefficient, p is the estimated parameter variance, e is the allowable error range; in step S3, Monte Carlo random sampling is performed and combined with the spatial coordinate verification mechanism to eliminate abnormal points.

[0069] This application realizes dynamic sampling planning by setting up a sampling model formula combined with Monte Carlo simulation technology. It can optimize the sampling distribution of farmland operation units in a large range, and use Monte Carlo simulation to perform multiple iterative calculations on the randomness in the sampling process and statistically distribute the errors, so that the coverage and representativeness of the sampling points in different spatial positions are guaranteed to the greatest extent. At the same time, in the sampling stage, the minimum control strategy of the spatial block is used to exclude or re-distribute the areas with outliers or excessive heterogeneity, thereby ensuring the accuracy and stability of the data while taking into account the sampling efficiency, and providing a reliable data basis for subsequent farmland information analysis and decision-making. This setting reduces unnecessary data collection and analysis workload while ensuring monitoring accuracy, effectively reducing the overall time and cost investment.

[0070] In the example of the present application, the schematic diagram uses three sampling points collected in each minimum agronomic unit. The numbering can be simple 1, 2, 3, or 1-1, 1-2, 1-3 for multiple farmlands. According to the preset agricultural monitoring accuracy requirements and the principle of uniform sampling, the positions of these three sampling points are automatically generated in the minimum agronomic unit.

[0071] The present application sets up multiple sampling points within the minimum agronomic unit. The data of multiple sampling points can verify each other, reducing the errors and deviations that may be caused by a single sampling point. Multiple sampling points are arranged according to the principle of uniform sampling. The function of automatically generating sampling point locations improves monitoring efficiency and reduces the time and energy cost of manually arranging sampling points.

[0072] As a preferred example of the present application, in step S4, the flight route of the UAV is planned according to the sampling point locations of all the smallest agronomic units in the farmland. According to the optimal control strategy for flight efficiency between the monitoring sample points, for example, fewer turns, closest distance, consistent sampling azimuths, etc., the system can automatically generate a flight route that is both efficient and practical. Unlike traditional route planning methods, this solution does not require setting the route overlap rate, but only sets the flight altitude according to the accuracy requirements of the agricultural indicators.

[0073] As a preferred example of this application, in step S4, the flight route of the drone uses an improved ant colony optimization algorithm, sets the pheromone volatility coefficient ρ=0.1, the inspiration factor α=1, the expectation factor β=5, solves the multi-objective traveling salesman problem (TSP), and uses the shortest total flight distance, the least number of turns, and the consistent image azimuth as optimization constraints for intelligent generation. r =0.1, inspiration factor α =1, expectation factor β =5, construct a multi-objective optimization function:

[0074]

[0075] in is the weight coefficient, is the node spacing, is the number of turns, To preset the optimal collection azimuth, it should be kept parallel to the sowing row azimuth;

[0076] The ant colony algorithm is used to simulate the process of multiple "ant individuals" searching for paths in space, and the optimization of the path is guided according to the set heuristic factors and expected variance. After each ant completes a search, the distribution of pheromones is adjusted according to its path weight and quality, and the optimized flight options are gradually gathered. Finally, the algorithm solves the task according to the set quality constraints and realizes the optimal flight path planning. The system can automatically calculate and intelligently generate the most suitable drone routes for farmland monitoring, which maximizes flight efficiency and reduces time and energy consumption.

[0077] The route generated by this improved ant colony optimization algorithm in this application can not only minimize the number of sharp turns during flight while maintaining the shortest total flight distance, but also ensure a high consistency azimuth when collecting images, thereby shortening the flight distance and reducing energy consumption while reducing the mechanical loss and operating difficulty of the UAV body, and effectively improving the comparability of image data and the efficiency of subsequent stitching processing, and achieving rapid and accurate data collection in large-scale farmland or complex terrain conditions, thereby improving the efficiency of UAV field inspections and the quality of agricultural monitoring, and providing higher accuracy and consistency source data for subsequent image analysis and agronomic decision-making.

[0078] The drone agricultural monitoring aerial survey method based on spatial sampling optimization described in the present application makes the flight path of the drone between monitoring sample points more reasonable and efficient, reduces unnecessary turns and flight distances, thereby saving flight time and energy consumption. While ensuring monitoring accuracy, the monitoring tasks of all sampling points are completed with the optimal flight path, thereby improving monitoring efficiency and accuracy.

[0079] As a preferred example of the present application, the improved ant colony algorithm introduces a quantized pheromone attenuation mechanism. When the path pheromone concentration exceeds the threshold When Dynamic reset ( l is the recession rate, t is the number of iterations). This setting quantizes the pheromone attenuation, so the algorithm can find a better solution in fewer iterations and improve the optimization efficiency.

[0080] As a preferred example of the present application, in step S5, the drone automatically flies according to the set sampling route, obtains point-by-point images of the sampling points, and transmits back the image data, which is classified and sorted according to the smallest agronomic unit. In the specific example of the present application, the drone is equipped with a high-resolution camera. During the flight, when it reaches above each preset sampling point, the camera automatically triggers shooting to obtain the image data of the sampling point; after shooting, the drone transmits the image data back to the ground control station in real time through the built-in wireless communication module; after receiving the image data, the ground control station automatically identifies and determines the smallest agronomic unit to which the image data belongs based on the location information (such as GPS coordinates) contained in the image data; then, the ground control station classifies and stores the image data according to the smallest agronomic unit, establishes corresponding folders or database tables, and ensures that the image data of each smallest agronomic unit can be archived in an orderly and clear manner; in this way, users can easily find and browse the corresponding image data according to the smallest agronomic unit, providing strong support for subsequent farmland agricultural conditions analysis.

[0081] As a preferred example of the present application, the method for determining the preset flight altitude in step S5 includes:

[0082] Based on the spatial resolution requirements of the target agricultural parameters, the formula Calculate the flight altitude, where:

[0083] H is the flight altitude of the drone (unit: meter);

[0084] GSD is the ground sampling distance (unit: cm / pixel), which is set according to the monitoring accuracy requirements. GSD ≤2D (D is the minimum recognition size of the target crop feature);

[0085] f is the focal length of the camera (unit: mm);

[0086] p x is the camera pixel size (unit: micrometer);

[0087] The altitude control error does not exceed ±0.5%H, and matches the band penetration characteristics of the spectral sensor, meeting the signal-to-noise ratio requirement of the exponential inversion.

[0088] In this application, by using the above height calculation formula, the required spatial resolution requirements are first determined according to the crop growth characteristics or monitoring targets, and then the GSD and f divided by the camera pixel size p xThe most suitable flight altitude H is obtained in theory, and this altitude is matched with factors such as camera focal length, lens angle of view, and sensor sensitivity, so that the acquired image has stable optical imaging quality while ensuring resolution. It reduces the time and energy waste caused by excessive or insufficient flight while ensuring high-resolution acquisition, and effectively avoids the problems of uneven image clarity and data deviation caused by arbitrary setting of flight altitude in traditional methods. By introducing an allowable error of ±0.5%H and combining it with ground spectral calibration, environmental interference factors can be quickly eliminated during the monitoring process to improve data reliability and consistency. Especially when facing large areas of farmland or areas with complex terrain, this solution can dynamically adjust the flight altitude according to the needs of different crop feature recognition, ensuring that each shot can obtain key agricultural information with the best imaging ratio, which not only reduces the difficulty of subsequent stitching or feature recognition, but also significantly reduces the burden of data analysis.

[0089] As a preferred example of the present application, in step S6, the sampling statistics method adopts an adaptive sampling execution method, according to the crop growth heterogeneity coefficient d = s / m Dynamically adjust sampling density ( d ≥0.3 n ≥5,0.1≤ d <0.3≤ n <5, d <0.1 n = 1, and the spatial representativeness was verified by the Moran index (∣ I ∣≤0.2).

[0090] The statistical sampling principle in step S6 is specifically:

[0091] Spatial stratified sampling is performed for each minimum agronomic unit, and the heterogeneity coefficient of crop growth within the unit is used to determine the spatial distribution of the crop growth. d = s / m ( s is the standard deviation of historical data, m is the mean) to dynamically adjust the sampling density to meet:

[0092] when d ≥0.3, high-density sampling mode is adopted, and the number of sampling points n ≥5;

[0093] When 0.1≤ d <0.3, medium density sampling mode is adopted, 3≤ n <5;

[0094] when d When <0.1, a low-density sampling mode is used. n =1;

[0095] The sampling process verifies the representativeness of the sample through spatial autocorrelation analysis to ensure the Moran index I ∈[-0.2,0.2], achieving the Pareto optimality of sampling error and flight energy consumption.

[0096] This application establishes the crop growth heterogeneity coefficient d = s / m Dynamic evaluation mechanism based on the standard deviation of farmland historical data s With the mean m To measure the discreteness of crop growth d = s / m The δ value is divided into three levels: high heterogeneity, medium heterogeneity and low heterogeneity. Different levels correspond to different numbers of sampling point layouts. By arranging 5 sampling points in high heterogeneity areas, 3 to 4 sampling points in medium heterogeneity areas, and 1 to 2 sampling points in low heterogeneity areas, stratified monitoring and sampling of crop growth conditions can be achieved. In order to further improve the accuracy and representativeness of sampling, the Voronoi diagram can be used for spatial optimization. By dividing the farmland into several polygonal units, each sampling point can cover the crop area closest to it, thereby effectively capturing the differentiated information of crop growth with fewer point layouts. This application obtains the standard deviation s and mean m Then, the level of heterogeneity is determined according to the threshold, and the location of the sampling points is accurately determined by combining the segmentation method of the Voronoi grid. Finally, a set of sampling strategies that dynamically adapt to the distribution characteristics of crop growth is formed. d The real-time calculation and evaluation of values ​​can flexibly adjust the number and location of sampling points at different growth stages of crops, and generally achieve accurate control of farmland conditions and rational allocation of resources, laying a refined foundation for subsequent farmland data collection and analysis.

[0097] As a preferred example of the present application, in step S6, various algorithms are used to extract agricultural information of the smallest agronomic unit from the image, and provide information for making farmland management decisions. After the agricultural conditions of all the smallest agronomic units are summarized, the agricultural information of the entire farmland is obtained. In the example of the present application, one or more image analysis algorithms suitable for farmland agricultural condition monitoring are selected, such as support vector machines (SVM) and random forests in machine learning algorithms, or convolutional neural networks (CNN) in deep learning algorithms, and features in the image are automatically learned and classified or regressed for prediction.

[0098] The drone agricultural condition monitoring and aerial survey method based on spatial sampling optimization described in this application, for field production, uses drones to collect agricultural condition information data, divides the farmland into multiple minimum agronomic units through sampling strategies, and sets multiple sampling points in each operation unit. According to the sampling points in all operation units, the optimal flight route is planned, and the drone flies along the route to collect image data in each operation unit, and uses computer science to analyze the images to obtain the agricultural condition of the minimum agronomic unit, and finally provides decision support for farmland management. In this way, the traditional method of comprehensive coverage of farmland areas is changed to monitoring key points in the minimum agronomic unit, which greatly narrows the scope of farmland monitoring, reduces unnecessary data collection, and greatly reduces flight time and the amount of collected data, thereby improving flight efficiency, reducing the time-consuming image stitching, and can achieve rapid and accurate diagnosis of farmland agricultural conditions after flight, which conforms to the basic model of farmland management. The drone agricultural condition monitoring aerial survey method based on spatial sampling optimization described in the present application solves the contradiction between improving drone monitoring accuracy and improving monitoring efficiency through the collaborative design of multi-objective ant colony algorithm optimization route and dynamic stratified sampling mechanism. It is suitable for precise and rapid monitoring of agricultural conditions in field crop scenes, optimizes the data collection efficiency of drones, greatly improves the accuracy and timeliness of monitoring, and significantly reduces costs, thus achieving the purpose of high-precision and high-timeliness drone monitoring of agricultural conditions. Specific embodiments

[0100] The present application discloses a method for drone agricultural monitoring based on spatial sampling optimization, comprising:

[0101] 1. Geographical spatial modeling and unit division

[0102] Based on the spatial registration of satellite remote sensing images and drone orthophotos, a digital farmland base model with centimeter-level accuracy is constructed;

[0103] According to the topological structure of the irrigation system and the agricultural machinery operation zones, the monitored plots are divided into a set of minimum agronomic units with independent agronomic attributes to form a GIS topological structure.

[0104] 2. Adaptive Sampling Model Construction

[0105] Establishing the coefficient of heterogeneity of crop growth d = s / m Dynamic assessment mechanism to perform spatially stratified sampling according to heterogeneity levels;

[0106] Highly heterogeneous regions ( d ≥0.3): ≥5 sampling points are arranged, and Latin hypercube sampling is used to ensure spatial balance;

[0107] Medium heterogeneous area (0.1≤ d<0.3): Arrange 3-4 sampling points and use Voronoi diagram to optimize the minimum spacing;

[0108] Low heterogeneity region ( d <0.1): set up one representative sampling point;

[0109] The Moran index verifies the independence of the sample space (∣ I ∣≤0.2), achieving precision control of sampling error ≤5%.

[0110] 3. Intelligent route planning

[0111] Using the improved ant colony optimization algorithm to set the pheromone volatility coefficient r =0.1, inspiration factor α =1, expectation factor β =5, construct a multi-objective optimization function:

[0112]

[0113] Introducing the quantized pheromone update strategy, when the path pheromone concentration exceeds the threshold Dynamically reset to avoid local optimum.

[0114] 4. Altitude-precision coupling control

[0115] Based on the accuracy requirements of agricultural parameter inversion, the flight altitude is calculated using the optical diffraction model:

[0116]

[0117] in GSD ≤2D (D is the target feature size), multi-spectral sensor band adaptation error ≤3nm;

[0118] Adaptive sampling execution: Based on the crop growth heterogeneity coefficient Dynamically adjust sampling density ( d ≥0.3 n ≥5,0.1≤ d <0.3≤ n <5, d <0.1 n = 1, and the spatial representativeness was verified by the Moran index (∣ I ∣≤0.2).

[0119] By adopting the drone agricultural monitoring aerial survey method based on spatial sampling optimization described in the present invention, it has the following beneficial effects compared with the traditional method:

[0120] 1: Optimization of the number of images acquired;

[0121] Traditional methods use full coverage images, which require the collection of all image data in the area (e.g., 2,000+ images are required for 1,000 mu of farmland), resulting in a data redundancy rate of up to 70%-80%. Under high-precision requirements (e.g., GSD≤2cm), the number of images is inversely proportional to the square of the resolution, and the amount of data increases exponentially (e.g., if the resolution is doubled, the number of images increases by 4 times).

[0122] The drone agricultural monitoring aerial survey method based on spatial sampling optimization described in the present invention acquires only key point images through directional collection at sampling points based on a dynamic sampling model, reducing the number of images by 60%-80% (e.g., the number of images for an area of ​​1,000 mu is reduced from 2,000 to 400-800).

[0123] Specific implementation example data: In a 200-acre monitoring area, the number of images dropped from 1,800 (full coverage) to 400 (sampling), and the amount of data was reduced by 77.8%.

[0124] In addition, the UAV agricultural monitoring aerial survey method based on spatial sampling optimization described in the present invention also realizes the function of spatial distribution optimization.

[0125] Through the Moran index (∣ I ∣≤0.2) Ensure the spatial independence of sampling points to avoid repeated coverage. The Voronoi diagram constrains the spacing between adjacent points to be ≥3 times the image coverage diameter (for example, if the coverage diameter of a single image is 5m, the point spacing is ≥15m), further reducing redundant images.

[0126] The method for agricultural condition monitoring and aerial surveying by unmanned aerial vehicle based on spatial sampling optimization described in the present invention achieves the following effects in terms of the number of images obtained:

[0127] (1) Reduction of image storage costs: The data volume of a 1,000-acre area is reduced from 500GB (full coverage) to less than 100GB.

[0128] (2) Improved transmission efficiency: Data transmission time on 4G networks is shortened from 3 hours to 40 minutes.

[0129] 2: Image analysis and processing time compression;

[0130] The traditional method uses image stitching and orthorectification. It takes 2-3 hours to stitch 1,000 images (still takes 30-60 minutes after GPU acceleration), and the accuracy is affected by lighting and terrain (error ≥ 5%). Full-area analysis based on orthophotos (such as NDVI calculation) requires processing hundreds of millions of pixels, and a single analysis takes 1-2 hours.

[0131] The present invention adopts a new method of agricultural monitoring aerial survey by drone based on spatial sampling optimization, which only processes the images of the sampling points independently, eliminating the stitching and orthorectification steps, and reducing the analysis time by more than 90% (in the embodiment, the 200-acre area is reduced from 2 hours to 10 minutes), achieving the purpose of direct analysis without stitching. And by using a cropped version of the ResNet-18 model (the number of parameters is compressed from 11M to 3M), the processing time of a single image is reduced from 0.5s to 0.1s. The parallel processing framework (such as CUDA acceleration) supports the simultaneous analysis of 100+ images, and the total processing time of 400 images is ≤1 minute, achieving the purpose of lightweight model acceleration.

[0132] The impact analysis efficiency comparison between the traditional method and the new method is as follows:

[0133]

[0134] 3: Reduced complexity of agricultural monitoring

[0135] The sources of complexity in traditional agricultural monitoring methods include:

[0136] Data preprocessing is complex: problems such as image registration error, uneven lighting, and stitching gaps need to be solved;

[0137] The whole area has a high computing load: it needs to process a large amount of pixel data (for example, a 1,000-acre orthophoto contains 10^9 pixels), and the algorithm complexity is O(N²);

[0138] Redundancy of decision-making information: The data of the entire region contains a large amount of irrelevant details, and additional noise filtering is required.

[0139] The present invention adopts a new method of agricultural monitoring by drone based on spatial sampling optimization, and adopts stratified sampling to reduce the dimension: the monitoring problem is reduced from "pixel-level analysis of the whole area" to "statistical inference of limited sampling points", and the algorithm complexity is reduced from O(N²) to O(k) (k is the number of sampling points, k N).

[0140] Specific implementation data: Using the new method of the present invention, the number of pixels analyzed in a 200-acre area is reduced from 200 million to 400 sampling points, and the amount of calculation is reduced by 99.998%. By dynamically focusing on key areas, based on the heterogeneity coefficient d ( ) Dynamically allocate resources: When the heterogeneity coefficient is located in the high heterogeneity area ( d ≥0.3), dense sampling is used to capture details; when the heterogeneity coefficient is in the low heterogeneity region ( d <0.1) to avoid over-calculation.

[0141] The new method of agricultural condition monitoring and aerial surveying by UAV based on spatial sampling optimization described in the present invention can directly output the statistical results of the minimum agronomic unit level (such as mean ± variance), replace the traditional full-area heat map, reduce decision-making interference information, and make the decision logic clearer.

[0142] The complexity comparison between the traditional method and the new method is as follows:

[0143]

[0144] 4: Comprehensive advantages in high precision and large area scenarios

[0145] In the case of large farms with an area of ​​more than 10,000 mu and centimeter-level monitoring accuracy requirements, the advantages of the new method of agricultural monitoring by drones based on spatial sampling optimization are further magnified:

[0146] (1) Hardware resource saving:

[0147] The drone’s endurance requirement is reduced from 6 hours (full coverage) to 2 hours (sampling route);

[0148] The computing server configuration can be reduced from a 32-core GPU to an 8-core CPU to meet real-time analysis requirements.

[0149] (2) Real-time monitoring capability:

[0150] Supports the completion of the "aerial survey-analysis-decision-making" closed loop within 3 hours (traditional methods take 1-2 days).

[0151] (3) Enhanced scalability:

[0152] Through distributed sampling (block parallel aerial survey), it supports ultra-large-scale farmland monitoring of millions of acres, and the system throughput increases linearly. Combined with the drone airport system, it can greatly increase the monitoring area of ​​a single airport system and improve the system efficiency.

[0153] Taking 5,000 mu of standardized farmland as an example, the cost of the traditional method of agricultural monitoring and the new method of this application are compared:

[0154]

[0155] The UAV agricultural monitoring aerial survey method based on spatial sampling optimization proposed in this application reduces the number of image acquisitions through dynamic sampling, reduces processing time through intelligent route optimization, and simplifies monitoring complexity through layered analysis. It can achieve the following in scenarios with large monitoring areas and high precision requirements:

[0156] (1) The amount of image data is reduced by 60%-80%, and the storage and transmission costs are significantly reduced;

[0157] (2) Processing efficiency is improved by 98%, supporting near-real-time agricultural decision-making;

[0158] (3) The system complexity is reduced by more than 90%, adapting to the deployment of edge computing devices;

[0159] (4) The overall cost is reduced by 75%, making it worthy of large-scale commercial promotion.

[0160] The drone agricultural monitoring aerial survey method based on spatial sampling optimization described in this application redefines the technical paradigm of drone agricultural monitoring, shifting from "full acquisition + post-processing" to "precise sampling + real-time analysis", providing an efficient and scalable technical foundation for smart agriculture.

Claims

1. A method for agricultural monitoring and aerial surveying by unmanned aerial vehicle based on spatial sampling optimization, characterized in that: The steps include: S1: Pre-build the GIS model of the monitoring plots, and draw the plot boundaries and the topological structure of the minimum agronomic unit based on the geographic coordinate information; S2: Establish a sampling model based on the agricultural monitoring accuracy index, and calculate the optimal number of sampling points for each minimum agronomic unit based on the minimum agronomic unit area parameter and crop growth heterogeneity index; S3: Execute the Monte Carlo random sampling algorithm to generate a set of monitoring sampling points that meet the preset number and conform to the spatial distribution constraints, and exclude abnormal sampling points that exceed the unit boundary or have overlapping areas; S4: Apply the traveling salesman problem optimization algorithm to generate global route planning, with the minimum number of flight turns, the shortest total flight distance and the consistent image acquisition azimuth as multi-objective optimization constraints, and intelligently generate drone routes; S5: Implement automated control of aerial survey tasks, collect fixed-point images at preset altitudes based on RTK positioning technology, and build a metadata indexing system with temporal and spatial correlation; S6: Deploy a deep learning image analysis model, analyze crop parameters through feature extraction technology, use sampling statistics to obtain the smallest agronomic unit information, and integrate to generate plot-level monitoring reports.

2. The method for agricultural condition monitoring and aerial surveying by unmanned aerial vehicle based on spatial sampling optimization according to claim 1 is characterized in that: In step S1, a digital farmland base model with centimeter-level accuracy is constructed through spatial registration of satellite remote sensing images and drone aerial survey orthophotos.

3. The method for agricultural condition monitoring and aerial surveying by unmanned aerial vehicle based on spatial sampling optimization according to claim 1 is characterized in that: The minimum agronomic unit is a spatial management unit with independent agronomic management characteristics.

4. The method for agricultural condition monitoring and aerial surveying by unmanned aerial vehicle based on spatial sampling optimization according to claim 3 is characterized in that: The division method of the minimum agronomic unit includes at least one of an irrigation system topology structure and an agricultural machinery operation partition.

5. The method for agricultural condition monitoring and aerial surveying by unmanned aerial vehicle based on spatial sampling optimization according to claim 1 is characterized in that: In step S2, the sampling model of the sampling points satisfies ,in n is the number of sampling points, Z is the confidence coefficient, p is the estimated parameter variance, e is the allowable error range.

6. The method for agricultural condition monitoring and aerial surveying by unmanned aerial vehicle based on spatial sampling optimization according to claim 1 is characterized in that: A spatial coordinate verification mechanism is established. In step S3, Monte Carlo random sampling is performed and combined with the spatial coordinate verification mechanism to eliminate abnormal points.

7. The method for agricultural condition monitoring and aerial surveying by unmanned aerial vehicle based on spatial sampling optimization according to claim 1 is characterized in that: In step S4, an improved ant colony optimization algorithm is used, the pheromone volatility coefficient ρ=0.1, the heuristic factor α=1, the expectation factor β=5 are set, a multi-objective optimization function is constructed, and the multi-objective traveling salesman problem is solved. The multi-objective optimization function is: is the weight coefficient, is the node spacing, is the number of turns, To preset the optimal collection azimuth, keep it parallel to the sowing row azimuth.

8. The method for agricultural condition monitoring and aerial surveying by unmanned aerial vehicle based on spatial sampling optimization according to claim 7 is characterized in that: In step S4, the improved ant colony algorithm introduces a quantized pheromone attenuation mechanism. When the path pheromone concentration exceeds the threshold When Dynamic reset, λ is the recession rate, t is the number of iterations.

9. The method for agricultural condition monitoring and aerial surveying by unmanned aerial vehicle based on spatial sampling optimization according to claim 1 is characterized in that: In step S5, the method for determining the preset flight altitude includes: Based on the spatial resolution requirements of the target agricultural parameters, the formula Calculate the flight altitude, where: H is the flight altitude of the drone, in meters; GSD is the ground sampling distance, in centimeters / pixel, which is set according to the monitoring accuracy requirements. GSD ≤2 D , D The minimum recognition size of the target crop feature; f is the focal length of the camera, in millimeters; p x is the camera pixel size in microns; The altitude control error does not exceed ±0.5%H.

10. The method for agricultural condition monitoring and aerial surveying by unmanned aerial vehicle based on spatial sampling optimization according to claim 1, characterized in that: The principle of the sampling statistical method in step S6 is: Spatial stratified sampling is performed for each minimum agronomic unit, and the heterogeneity coefficient of crop growth within the unit is used to determine the spatial distribution of the crop growth. δ = σ / μ , σ is the standard deviation of historical data, μ As the mean, the sampling density is adjusted dynamically: when δ ≥0.3, high-density sampling mode is adopted, and the number of sampling points n ≥5; When 0.1≤ δ <0.3, medium density sampling mode is adopted, 3≤ n <5; when δ When <0.1, a low-density sampling mode is used. n =1; The sampling process verifies the representativeness of the sample through spatial autocorrelation analysis to ensure the Moran index I ∈[-0.2,0.2].

Images